Node Your Homework presents: a paper I wrote with a classmate for a cognitive psychology class, posted here in slightly modified form, with my co-author's permission.

Theories of Attention

9 October 1998 (updated 30 May 2003)
by fuzzy and blue and an anonymous classmate at Harvey Mudd College

Attention has been of interest to psychologists ever since William James first wrote about it in his Principles of Psychology in 1890. However, attention is a complicated phenomenon, one that is difficult to define and discuss, much less observe. Because early 20th century psychology was dominated by behaviorist theory, which emphasized observable facts, it took more than 60 years for the first modern theory of attention to develop. E. C. Cherry reawakened interest in theories of attention when he published the results of his investigation of what he called "the cocktail party problem" in 1953 (Hampson & Morris 112). Subsequent theories of attention have since expanded and refined Cherry's ideas to considerably higher levels of sophistication, although the problem of attention is far from a closed one. In this paper, we will be examining the evolution of theories of attention, starting with the results of Cherry's "dichotic listening" experiments and continuing on to modern cognitive theories of divided attention and attentional resources.

Cherry's 1953 article "Some Experiments on the Recognition of Speech, with One and with Two Ears" presented the results of several experiments using the "dichotic presentation" technique, in which subjects listening to two different messages at once (one in each ear, through headphones) were able to "shadow" one message, repeating it as they listened, but retain little to no information about the non-attended message (Komatsu 34). In fact, listeners were, for the most part, not even able to recognize when the non-attended message was played backwards, or altered in some other way (Hampson & Morris 113). Although Cherry's article presented an interesting new technique for experimenting with attention, and suggested a connection between focused attention and the physical features of material that was being attended to, it did not attempt an explanation of his results. As a result, later research that made use of dichotic listening tasks was often more focused on the effects of the task than in the cognitive processes involved in attending to one set of stimuli over another (113). Although Cherry's work paved the way for future research in attention, an actual theory describing its mechanisms and processes had yet to be developed.

Donald Broadbent proposed the first comprehensive model of attention in 1958, based on the work of Cherry and others. According to Broadbent's filter theory, the first of the so-called "bottleneck" theories of attention, people can attend to only one message at a time. Attention was modeled as a physical channel that could allow only one message --- identified by its physical characteristics --- to pass through into the processes of perception and short-term memory. Non-selected (unattended) messages were supposedly stored in a temporary memory buffer where soon only their physical properties remained available for retrieval (Komatsu 41-2). In this way, the model accounted for the fact that participants in Cherry's dichotic listening experiments could recall basic physical features of the unattended messages, such as pitch and loudness, or whether they were listening to a voice or a steady tone. The other important feature of Broadbent's filter theory is that it clearly defined the idea of "attending to a message" as "allowing it [the message] into consciousness" (42). However, Broadbent's theory could not explain why certain words, such as a subject's name, seemed to pierce the selective filter, a phenomenon documented in 1959 experiments by N. Moray. By 1960, experimental evidence had also failed to support Gray and Wedderburn's filter theory-based prediction that subjects would be unable to identify a two-syllable word if each syllable was presented through a different aural channel (Hampson & Morris 115). Nor could the Broadbent model account for Anne Treisman's discovery of "intrusions" into attention, during which listeners accidentally followed a message from one ear to another (Treisman 45-6). Broadbent's single-filter, limited-capacity model of attention clearly had its limitations.

Anne Treisman's article, "Contextual Cues in Selective Listening," exposed many of the flaws with Broadbent's single-filter, limited-capacity model of attention. However, rather than completely abandon her predecessor's ideas, Treisman chose to adapt them somewhat to explain her results (47-8). These adaptations were subsequently incorporated in her own theory of attention. Instead of a single filter, Treisman's model had a hierarchy of tests, including one for physical properties of words, and others for syllable patterns, grammar, and finally semantic meaning. As they ran the gauntlet of filters, messages that failed various tests became attenuated to lessen interference with the selected message, but were not discarded until the final filter, which examined the semantic meaning of the messages. Hence an unattended message containing the subject's name or an emotionally charged word would get to the semantic filter and would be automatically responded to when semantically processed (Hampson & Morris 115-6). The shortcomings of this theory are that its explanation of an attenuated message eliciting an "automatic" response is somewhat contrived. Treisman does not explain how this attenuated message is responded to. It would seem, then, that Treisman's attempt to account for the response to emotionally charged words in a filter theory model of attention was unsuccessful.

The next widely accepted model and the last of the major bottleneck theories of attention was proposed in 1968. Anthony and Diana Deutsch and Donald Norman further altered Treisman's adaptation of Broadbent's theory by suggesting that all incoming information is semantically processed before being filtered out; in effect, this new theory swapped the filtering process and the data intensive information processing procedures. Some supporting evidence of this theory is that the definition of the word "bank" in the sentence "They threw rocks at the bank yesterday" can be swayed by introducing that definition into the unattended ear (116-7). The Deutsch-Norman handles emotionally charged words much better than the two previous theories but it does this at the expense of a huge amount of semantic processing. In general, the bottleneck theories are based on a stage theory of attentional processing that makes assumptions that modern theories have criticized.

D. Kahneman introduced the first of the modern attentional resource theories in 1973. Resource theories explain attentional limitations as a limit on the capacity of the mind instead of as a physical filtering process. Kahneman's theory states that there is a finite attentional capacity of the mind that is controlled by "enduring dispositions," "momentary intentions," and "arousal." It also predicted that, as a skill becomes more automated, it becomes more streamlined and takes up less of one's attentional capacity (119-20). What his theory does not address is people's ability to divide their attention and concentrate on two or more tasks at the same time. In order to explain divided attention, D. A. Allport suggested that the limitations of attention were not with the capacity of the information holder but with the mechanisms that process the information. These mechanisms are separate for different types of inputs, i.e. there may be different mechanisms for sound input and for visual input (121-2). Many other capacity theories suggest a more complex version of Allport's theory such that, instead of one capacity space for incoming information, there are several (this explanation accounts for some confusing data regarding which types of activities interfere with each other). However, as of the writing of our sources, no one of these theories seem to have gained widespread acceptance in the cognitive psychological world. As Hampson and Morris state, "the jury is still out" (113).


Hampson, Peter J., and Peter E. Morris. Understanding Cognition. Cambridge: Blackwell Publishers, Inc. 1996.

Komatsu, Lloyd. Experimenting With the Mind: Readings in Cognitive Psychology. Pacific Grove, CA: Brooks/Cole Publishing Company. 1994.

Treisman, Anne. "Contextual Cues in Selective Listening." Quarterly Journal of Experimental Psychology. 12, 1960, pp. 242-48. Reprinted in Experimenting With the Mind: Readings in Cognitive Psychology.

Update, 10 April 2012: Please check out quintopia's writeup, below, for much more (and more recent) information on this subject.